Non-isotropic voxels



Dear MRtrix experts,

I’m writing to ask if it’s better to interpolate first and then use isotropic voxels to run ACT, or I should put the size of voxels in the header information. Does the MRtrix3 automatically do interpolation for you in that way?

Thank you in advance,



Hi Yixin,

The size of the (anisotropic) voxels should already be in the image header based on however the data were initially obtained.

The tractography code in MRtrix3 all operates in “scanner space” / “real space”, acquiring interpolated values from any input images. So having a 5TT image with anisotropic voxels should work seamlessly, and hence there shouldn’t be any reasonable justification for resampling the data onto an isotropic grid first.



Hi Rob,

Thank you for the reply.

You mentioned that I should use the 5TT image for tckgen, I would like to ask if I -include, -seed, and -mask [white matter mask], will that be similar to input 5TT image? I know in the case of 5TT image, the seed will then be generated from the WM/GM interface, instead of randomly sampled from the white matter mask. I’m wondering if there are any other differences that I should be aware of (e.g., the fiber tracking will be restricted to white matter).